It is different at the orthogonal to the speed of light ;)

When I was working in model validation, I always was spending a lot of time examining models' business aspects and trying to assess the monetary impact of the model's shortcomings. In general, risk models can be divided into two groups: capital models and business models.

The former category covers models used to calculate potential losses in order to estimate the capital requirements. They usually involve scenario analysis and stress testing, and are subject to the regulatory scrutiny - they are often called regulatory models. These models try to balance regulatory expectations and the bank's risk appetite. Sometimes they are constrained by the regulator's acceptance of modeling techniques. This often results in utilizing basic techniques like linear regression.

The second category is about optimizing the business, i.e. maximizing profits. Here, the loose mainly equals profit not made. Virtual any technique is allowed if demonstrated correct. Machine learning and "artificial intelligence" may extend their claws ;)

From the model validation perspective, the second category is definitely more interesting, but few validators are able to handle the unfamiliar matter. Many of the solutions used were in-house designed and hence no academic literature is available.

And at the orthogonal to the speed of light is still different ;)